Task-Based Design of Modular Robots: Evolutionary Approach

Abstract

In an attempt to solve the problem of finding a set of multiple unique modular robotic designs that can be constructed using a given repertoire of modules to perform a specific task, a novel synthesis framework is introduced based on design optimization concepts and evolutionary algorithms to search for the optimal solution. The discrete optimization procedure is based on the use of assembly incidence matrix to represent modular robotic designs. Fitness function is formulated for each task separately to incorporate task-specific performance evaluation criteria. The fitness of every design is measured in simulation. Solution evaluation can be carried out in parallel in order to reduce synthesis time, because evaluating a certain design is independent of evaluating other designs. The feasibility of this approach is demonstrated by several examples.